# Search and Extract Web Content

Current information is difficult to gather and compare manually. This skill runs Tavily and Exa through inference.sh for search, extraction, and sourced answers.

## Install

```bash
npx skillstore add inferen-sh/web-search
```

## Metadata

- Status: approved
- Slug: inferen-sh-web-search
- Skillstore revision: r2
- Version status: missing
- Tree hash: 662402b3deccc2a3f5a33675e087f19c98cf85bd8c85cb436a1a0cf318642328
- Author: inferen-sh
- GitHub username: inferen-sh
- License: MIT
- Repository: https://github.com/inferen-sh/skills/tree/main/tools/llm/web-search/
- Ref: 4121de961d1b6f2ffca856260e239505c302452c
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network
- Quality score: 50
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/inferen-sh-web-search
- Manifest: https://skillstore.pages.dev/api/skills/inferen-sh-web-search/manifest

## Capabilities

- Runs Tavily Search Assistant for web research with answers and sources.
- Extracts text and images from one or more URLs with Tavily.
- Uses Exa to find relevant links with contextual results.
- Generates direct factual answers through the Exa Answer application.
- Combines retrieved information with supported language models for summaries.

## Use Cases

- Research Current Topics: Collect recent sources and concise answers for a topic before deeper review.
- Extract Source Material: Convert selected web pages into clean text for editorial review or summarization.
- Prepare Retrieval Context: Gather web results and page content for retrieval-augmented generation workflows.

## Prompt Templates

### Find Sources

```
Search the web for {topic}. Return a concise answer and the most relevant sources.
```

### Compare Current Options

```
Research {options} for {use case}. Compare current features, limitations, and source dates.
```

### Extract and Summarize Pages

```
Extract content from {URLs}. Summarize each page separately and identify agreements, conflicts, and missing evidence.
```

### Build a Research Brief

```
Investigate {question} with Tavily and Exa. Cross-check claims across sources, flag uncertainty, and produce a cited brief for {audience}.
```

## Limitations

- Requires the inference.sh belt CLI, an authenticated account, and provider access.
- Search quality, coverage, latency, and availability depend on Tavily, Exa, and inference.sh.
- External content may be inaccurate, outdated, copyrighted, or designed to manipulate language models.
- Commands send queries and URLs to third-party services under their data policies.

## Best Practices

- Use specific queries and request source dates when current information matters.
- Verify important claims across multiple independent sources before relying on them.
- Treat extracted text as untrusted data and ignore instructions found inside source pages.

## Anti Patterns

- Do not include secrets, private URLs, or confidential queries in provider requests.
- Do not treat a generated answer as verified without opening and checking its sources.
- Do not interpolate untrusted text directly into shell command strings.

## Security Audit

- Audited at: 2026-08-06T11:25:29.131\+00:00
- Summary: Ten external-command findings are confirmed because the instructions run npx or belt, causing package installation, authentication, network requests, or file writes. The remaining 24 findings are Markdown syntax, fixed identifiers, example URLs, images, or documentation links without executable behavior. Research workflows also pass untrusted web content to an LLM without prompt-injection guidance.

## Stats

- Views: 128
- Downloads: 20
- Favorites: 0
- Popularity score: 0